8. Configuration file parameters for the surveillance project - HopkinsIDD/gavi_vimc_cholera GitHub Wiki

A typical config file for the surveillance project would look like the following (taking the testing run for COD as an example):

runname: '202110gavi-3'
country: 'COD'
scenario: 'campaign-default'
clean: TRUE
incid:
  num_samples: 50
  redraw: FALSE
  use_country_incid_trend: FALSE
vacc:
  targeting_strategy: 'threshold_unconstrained'
  vac_incid_threshold: 0.001
  vac_unconstrained: TRUE
  vac_admin_level: 'admin2'
  vac_coverage: 0.68
  num_skip_years: 3
  vac_interval: 1
  sim_start_year: 2022
  vac_start_year: 2022
  vac_end_year: 2030
  sim_end_year: 2035
  use_mean_ir: TRUE
  mean_ir_span: 5
surveillance_scenario:
  surveillance_scenario: 'district-estimate'
  testing_sensitivity: 1
setting:
  incidence_rate_trend: FALSE
  outbreak_multiplier: FALSE
  random_seed: 1
optimize:
  save_intermediate_raster: TRUE
  save_final_output_raster: FALSE
  ir_pre_screening: FALSE

The followings are the definitions for all the parameters:

• runname: the unique name that's specific to each version of the Montagu server from which we download the demographic data, for this project however, it's 202302_survms when it comes to the location of the configs
• country: the 3-digit ISO code for a certain country
• scenario: either 'campaign-default' or 'no-vaccination' depending on which scenario to simulate
• clean: whether to delete the pre-existing final model output when new files with the same names are being generated during a new simulation
• incid:
  • • num_samples: how many stochastic runs to simulate/how many layers of raster files to carry around and get calculated each time
  • • redraw: whether to re-generate and renew the initial multi-layer (#num_samples) incidence rate rasters
  • • use_country_incid_trend: whether to use the country-level incidence rate trend for this specific country (will not be used if incidence_rate_trend is FALSE)
• vacc:
  • • targeting_strategy: it can be 'threshold_unconstrained', 'affected_pop', or 'incidence' depending on the type of the model; for the surveillance project, however, it will be 'threshold_unconstrained' only
  • • vac_incid_threshold: it can be 0.001, 2e-04, or 1e-04 (10 per 10,000, 5 per 10,000, or 1 per 10,000)
  • • vac_unconstrained: it is always TRUE for the surveillance study model
  • • vac_admin_level: it can be 'admin1', 'admin2', or 'both' depending what administration level(s) to simulate each time; 'admin1' means province-based targeting and 'admin2' means district-based targeting
  • • vac_coverage: it is always 0.68, representing the actual vaccination coverage of population above 1 year old in an area after it gets targeted for the vaccination campaign
  • • num_skip_years: it's usually 3, meaning how many years the same administration unit should wait until its next vaccination campaign
  • • vac_interval: it's usually 1, meaning how many years the same country should wait until its next vaccination campaign; if it's 1, then it can get the vaccines in every sequent year if it meets the threshold
  • • sim_start_year: usually 2022, meaning from which year to start the simulation
  • • vac_start_year: usually 2022, meaning from which year the vaccination campaign will become available
  • • vac_end_year: usually 2030, meaning at which year the vaccination campaign will end
  • • sim_end_year: usually 2035, meaning at which year the simulation will stop
  • • use_mean_ir: usually TRUE, meaning whether use the mean incidence rate averaged across the previous years
  • • mean_ir_span: usually 5, meaning how many years in the past contribute to the mean incidence rate calculation
• surveillance_scenario:
  • • surveillance_scenario: it can be 'district-estimate' (decentralized testing), 'global-estimate' (centralized testing), or 'no-estimate' (clinical definition), representing different cholera case bacteriological confirmation capacities
  • • testing_sensitivity: the sensitivity rate of the tests performed by the labs
• setting:
  • • incidence_rate_trend: whether to use the temporal incidence rate trend
  • • outbreak_multiplier: whether to use the spatial-temporal outbreak multiplier
  • • random_seed: the random seed, it's usually 1 when the above two parameters are both FALSE unless it's specified in set_all_parameters.R
• optimize:
  • • save_intermediate_raster: whether to save the intermediate to make the simulation process memory-efficient
  • • save_final_output_raster: whether to save all the final output of raster files, including expected cases, vaccination proportion, susceptible proportion raster files
  • • ir_pre_screening: usually FALSE, whether to skip any countries for modeling if none of their districts/provinces will ever go above the targeting threshold in the future